Hierarchical clustering.

Color matrices show the hierarchical clustering of 2343 (A, UPN1-38) and 2769 (B, UPN 39-77) selected genes (see “Materials and methods”) of gene expression data obtained from 76 B-CLL patients. The hierarchical clustering algorithm used is based on the average-linkage method. Red areas indicate increased expression, and green areas indicate decreased expression. Each column represents a single B-CLL sample and each row a single gene. Tables beneath the heat maps detail the individual patients (UPN) and some of their clinical characteristics. Treatment intensity, 0; no previous treatment, 1; patients requiring minimal treatment, defined as less than 6 months of therapy in the years of follow-up, 2; patients requiring continuous chemotherapy or chemotherapy with 2 or more agents or regimens, 1. Dendrograms at the top show the degree to which each B-CLL patient is related to the others with respect to gene expression. Note the separation into 2 groups comprising 8 and 30 (A) and 12 and 26 patients (B), respectively. Clustering within the leftmost branches of the dendrograms was caused, in part, by the high expression of a large number of ribosomal and translation-associated genes (see arrows at the left side of the heat maps).